Package: psAve 1.0.1

Daijiro Kabata

psAve: Model-Averaged Propensity Scores Selected by Prognostic-Score Balance

Constructs a model-averaged propensity score as a convex combination of candidate propensity score models, with mixing weights selected on a simplex grid to optimize covariate or prognostic-score balance, implementing the method of Kabata, Stuart and Shintani (2024) <doi:10.1186/s12874-024-02350-y>. Prognostic scores follow Hansen (2008) <doi:10.1093/biomet/asn004>: outcome models are fit on untreated units only. The resulting score is designed to be supplied directly to the matchit() function of 'MatchIt' as a distance measure or to the weightit() function of 'WeightIt' as a propensity score, with balance assessment via 'cobalt'.

Authors:Daijiro Kabata [aut, cre, cph]

psAve_1.0.1.tar.gz
psAve_1.0.1.zip(r-4.7)psAve_1.0.1.zip(r-4.6)psAve_1.0.1.zip(r-4.5)
psAve_1.0.1.tgz(r-4.6-any)psAve_1.0.1.tgz(r-4.5-any)
psAve_1.0.1.tar.gz(r-4.7-any)psAve_1.0.1.tar.gz(r-4.6-any)
psAve_1.0.1.tgz(r-4.6-emscripten)
manual.pdf |manual.html✨
DESCRIPTION |NEWS
card.svg |card.png
psAve/json (API)

# Install 'psAve' in R:
install.packages('psAve', repos = c('https://kabajiro.r-universe.dev', 'https://cloud.r-project.org'))

Bug tracker:https://github.com/kabajiro/psave/issues

Pkgdown/docs site:https://kabajiro.github.io

On CRAN:

Conda:

causal-inferenceepidemiologymatchingpropensity-score

4.18 score 5 scripts 5 exports 20 dependencies

Last updated from:a1f27ed69f. Checks:9 OK. Indexed: yes.

TargetResultTimeFilesSyslog
linux-devel-x86_64OK238
source / vignettesOK212
linux-release-x86_64OK226
macos-release-arm64OK219
macos-oldrel-arm64OK188
windows-develOK144
windows-releaseOK160
windows-oldrelOK152
wasm-releaseOK133

Exports:psavepsave_criteriapsave_matchpsave_weightsimplex_grid

Dependencies:argclicobaltcpp11farverggplot2gluegridExtragtableisobandlabelinglifecycleR6RColorBrewerrlangS7scalesvctrsviridisLitewithr

Getting Started with psAve
Introduction | Installation | A matching workflow with the lalonde data | Step 1: Estimate the model-averaged propensity score | Step 2: Match on the averaged propensity score | Step 3: Assess balance, including prognostic-score balance | Step 4: Estimate the treatment effect | Interpreting the output | print() | summary() | plot() | FAQ | Doesn't using the outcome bias my analysis? | References
Rendered frompsAve.Rmdusingknitr::rmarkdown

Last update: 2026-07-02
Started: 2026-07-02

Method details and design decisions
Notation | The simplex grid and tie-breaking | Step 1: Candidate models | Step 2: $\gamma$ — the model-averaged prognostic score | Step 3: $\lambda$ — the model-averaged propensity score | criterion = "logloss" | criterion = "smd" | criterion = "ks" | criterion = "prog" (the default; the paper's "Prog (Ave)") | Vertex mode | Criterion vs. display conventions for the SMD denominator | Five documented fixes relative to the paper's reference code | Relation to other software | Limitations | References
Rendered frommethod-details.Rmdusingknitr::rmarkdown

Last update: 2026-07-02
Started: 2026-07-02

Reproducing the published IPW workflow
Introduction | ATT weighting | The ecosystem route: psave_weight() | The direct route: weights() | The paper's estimator: survey::svyglm() | ATE weighting | Note on estimand-specific formulas | References
Rendered fromweighting.Rmdusingknitr::rmarkdown

Last update: 2026-07-02
Started: 2026-07-02